The Journey to the Private Cloud: The Business Buy-in
Planetary Cloud: The Rise of the Billion-Node Machine

Virtualization as ‘Continuous Workload Consolidation’ Paradigm

Imported from http://consultingblogs.emc.com/ published May 16 2010

 

At EMC Consulting we are leading the charge to the Private Cloud for our Clients. This is starting with full-scale server and datacenter virtualization. Most clients are initially interested in the consolidation abilities of this technology to reduce their physical server footprint. Projects are drawn up and concrete plans made. Troops assembled, and then the ‘storming of the physical datacenter beginsCool’ (‘Once more unto the breach dear friends, once more!’ – Shakespeare’s Henry V – V for virtualization of course)

Amidst all this action, it occurred to me that the task of deep-range planning for organizations regarding where virtualization is headed, perhaps take second place, if it occurs at all.

Virtualization is considered at the server level as the ability to decouple and abstract an operating system from the physical hardware it resides on; encapsulated in a virtual machine. Yes, and you’re quite correct, that is simply one definition, but the one most people are familiar with. Keywords are ‘decouple’, ‘abstract’ and ‘virtual’.

However, did we actually ponder what all this actually means for the concern itself and its strategic business thinking? Let’s take a slightly different view of server virtualization. Not simply the ability to consolidate workloads (VMware parlance), but as the transformation of workloads, such that they are being continuously consolidated.

Continuous workload consolidation (CWC – for this blog only) is increasingly driving the multi-core joy we're experiencing in the processor market. It is one of the largest drivers for consolidation at the level of the CPU. This is a paradigm shift occuring over the last 2-3 years.

I believe the current score in the ‘Soccer Cores Super League – Currently Shipping Division’ is 12:8 to AMDParty!!!, with their ‘Magny-Cours’. Mind you, Intel’s ‘Nehalem-EX’ with 8-cores is not really that far behind - if at all.

EMC Consulting in Germany maintains a technology watch for our clients to indicate possible technology inflection points for senior executives that would like to basically know ‘what EMC is doing about it and how this affects the client's competitive landscape and their business strategy decisions’.

Look at what is going on in the background when we are talking cores:

  • Intel’s ‘Teraflops Research Chip with 80 simple cores working in unison shown in 2007
  • The ‘single-chip cloud computer with its 48 complex cores shown in 2009
  • There are others in the GPU market doing similar things, for example, chip designer ClearSpeed with 96-cores in the HPC market

Clearly all the other components of systems need to follow in leaps and bounds. We have solid state disks offering hundreds of thousands of I/Os per second, we have 10GigE, with 100GigE looming, and various Infiniband offerings providing even higher aggregate throughput as the High Performance Computing crowd has amply proven.

So lots of good stuff coming, and the IT Planners are rubbing their hands with glee! However, the CIO/CTO needs to think beyond simply some ‘tasty gear’.

The idea of CWC implies that rationalization of application/infrastructure architectures, desktop/server sprawl and operational efficiency should already be on the agenda today. This rationalization already provides the means to potentially reduce the physical real-estate not simply by a 10:1 or 20:1 ratio, but more like 100:1 or more (100 physical servers consolidated to a single physical server). Rationalization extends to the organization allowing rapid change to be implemented and value extracted accordingly (Business and IT agility are needed).

Doing some quick maths as an example: a current datacenter hosting 1000 physical machines, say 2U height servers in 42U racks, requires around 65 racks (18 servers per rack, although cooling may imply conservatively 16 servers per rack plus space for patch panels etc).

This does not include the racks to host physical network switches, SAN switches, and the 8,000 cables (say 8 cables per server, production/ management/ SAN/ iSCIS networks, KVM cables, and the list goes on…oops, forgot the patch panel cabling to the central switch complex – shall we double the number of cables?).

That is enormous! (by the way if you are struggling with that – EMC Consulting can help youWink).

The CWC concept with the technology advances coming allow this to be potentially reduced to a handful of physical servers (I mean like say 5-10 physical servers). However, this massive density of virtual machines on these servers requires the rationalization transformation mentioned earlier on the Cxx agenda.

That rationalization ensures that the organization is essentially ‘fit’ to use these advances as they become available. It ensures that current investments are focused to allow simple server replacement decisions to be made to provide that technology and move all the workloads to that new platform in seconds. This already provides further consolidation (we are running more workloads on less equipment rightGeeked)

These are big decisions we are talking about folks. Should we build a new datacenter now or not? Should we make do with what we have? What level of service can I offer with the resources in my possession? Do I need to offshore my IT? Do the current organizational structures and processes allow me to get this competitive advantage? Serious stuff.

When we start to factor in desktops and desktop applications this massively favours centralization of compute resource. However, the corresponding network technology, always-on, needs to be built out to ensure access to this brave new virtualization world. This would indicate the move to 10GigE, IPv6 and possibly fiber optics throughout should already form a large part of the infrastructure build-out that is required.

This is going to have many knock on affects. If we currently don’t know where our old physical server workload is actually running due to dynamic workload shifting (VMware DRS/DPM if we are watching the green IT profile) in the datacenter, then this virtualization of physical people presence through advanced networking may mean that a large part of the workforce operates from the home. After all, they’re just a phone call away. Indeed, with video conferencing advances, they can be easily contacted, information/workspaces shared/exchanged, and we can attach a face to the remote voice.

That raises a lot of questions - food for thought.

Do we need to have large office blocks anymore? Should an organization invest in real-estate on current scales? Will this change inner city landscapes and architectures? What about the distribution of small businesses and restaurants typically based in close proximity to these large commercial centers where thousands of employees operate? What is going to happen to the road congestion, indeed road and transport design? What about all those Tesco's stores built in the London financial districtsStick out tongue

Folks, there are many elements of our normal lives that are ultimately affected by the current server virtualization wave. Some holistic thought early on will allow pioneers of industry, creative geniuses, artists, urban designers, and hey, even you & I Wink to be able to heavily influence the ‘digital workplace’ of the not-too-distant future!!

Disclaimer

The opinions expressed here are my personal opinions. Content published here is not read or approved in advance by EMC and does not necessarily reflect the views and opinions of EMC.

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

The comments to this entry are closed.